www.nature.

com/scientificreports

OPEN Associations of interpersonal trust

with juvenile offending/conduct
disorder, callous‑unemotional
traits, and criminal recidivism
1,2* 3
Marcel Aebi  , Melanie Haynes  , Cornelia Bessler1,2 & Gregor Hasler  4

Interpersonal trust has been described as a core dimension of cooperative, mutually beneficial
interpersonal relationships but it is unclear if it is related to antisocial behaviours in youth. The present
study aimed at analysing a subsample of male juveniles who committed serious violent offenses
and met criteria of conduct disorder (JO/CD), and a subsample of healthy controls (HC) using a series
of trust games (TGs). Twenty-four male JO/CD and 24 age matched male HC performed a series of
eight one-shot TGs against different unknown human respectively computer opponents. Mixed
model analyses found a non-significant trend that JO/CD invested less points than HC during TGs. In
the subsample of JO/CD, the overall investment in TGs was found to be negatively associated with
self-reported uncaring behaviours and officially reported general re-offenses. Our findings suggest
some indication of an impaired ability of JO/CD to initiate mutually trusting relationships to others
that should be addressed in further research. Trust is a promising factor to predict general criminal
recidivism and can be a target for treatment of juveniles who committed violent offenses, for example
through the building of stable relationships to care givers. This study encourages future studies to
investigate the effects of trust-increasing psychosocial interventions.

The prevention of youth violence represent a significant public concern w ­ orldwide1. The cost of juvenile antisocial
behaviour is known to be high, particularly for those youth who become persistent offenders. In the 5th edition of
the American Psychiatric Association’s Diagnostic and Statistical Manual of Mental Disorders (DSM-5) and the
10th/11th revision of the International Classification of Diseases (ICD-10/-11) antisocial and violent behaviours
in youth is addressed by the diagnosis of conduct disorder (CD). A major criticism of the diagnostic criteria for
CD is, that they are based on behavioural symptoms and not on the underlying cognitive or emotional processes
that drive these s­ ymptoms2,3. Now, DSM-5 and the upcoming ICD-11 further allow to differ between subtypes
of CD according to the presence of “limited prosocial emotions” (referring to callous-unemotional traits that
includes a lack of empathy, a lack of remorse, shallow affect, or a lack of concern about performance). A large
body of research has investigated callous-unemotional traits, which were found to go along with a severe, violent,
and persistent pattern of antisocial behavior in children and a­ dolescents3,4.
Despite youth specialist forensic services having been introduced in many countries, and effective interven-
tions for antisocial youth being available, the treatment of juvenile offenders remains a challenge for many child-
and adolescent-psychologists and psychiatrists. The identification of additional useful treatment approaches for
changing delinquent behaviours/attitudes in order to impede criminal recidivism is necessary.
Interpersonal trust has been described as a core dimension of cooperative, mutually beneficial interpersonal
relationships and the basis for the development of a concept of fairness and moral identity in children and
­adolescents5,6. Trust can be defined as “a psychological state comprising the intention to accept vulnerability based
upon the positive expectations of the intentions or behaviour of another”7,p. 395. Interpersonal trust portrays a
social dilemma, because trusting another person is always associated with the risk of being cheated or exploited.
Although high levels of social trust seem beneficial for an individual, it would be a failure to trust anybody.

1
Research and Development, Corrections and Rehabilitation, Department of Justice and Home Affairs, Canton
of Zurich, Hohlstr. 552, 8090  Zurich, Switzerland. 2Department of Forensic Psychiatry, University Hospital of
Psychiatry Zurich/University of Zurich, Zurich, Switzerland. 3Translational Research Center, University Hospital
of Psychiatry and Psychotherapy Bern, Bern, Switzerland. 4Unit of Psychiatry Research, University of Fribourg,
Fribourg, Switzerland. *email: marcel.aebi@uzh.ch

Scientific Reports | (2022) 12:7594 | https://doi.org/10.1038/s41598-022-11777-6 1

Vol.:(0123456789)
 www.nature.com/scientificreports/

From a developmental perspective, interpersonal trust arises from the child’s attachment to caregivers in early
childhood. According to Erikson’s psychosocial ­theory8 children gain a sense of trust that the world is good dur-
ing their first year of life as they receive warm and responsive care. Children develop an “inner working model” on
trustful or unreliable relations based on the experiences of early interactions with primary ­caregivers9. In line with
these considerations, social trust was found to increase with age during childhood and ­adolescence10. Unsecure
bonding, inconsistent and harsh parenting and/or early child maltreatment affects the ability to develop future
trustful relationships to others and increase the risk of psychiatric d­ isorders5,11. Accordingly, low trust was found
to be associated with schizophrenia, borderline personality disorder and post-traumatic stress d ­ isorders12–14.
From a theoretical perspective, low interpersonal trust is also related to juvenile offending, CD, and CD-
associated callous-unemotional traits: First, the experience of early trustful relations to primary caregivers is
an important basis for the moral development of children and a­ dolescents5. Moral deficits were found to be
consistently related to antisocial behaviours and criminal recidivism in youths and ­adults15,16. Second, childhood
adversity and conduct problems were found to impact neural development for the recognition of others’ emotions
(see research on amygdala and responses to facial expression of anger)17–19 and the hostile attribution of others’
intentions in social interactions was found to drive aggressive b ­ ehaviour20,21. Third, the ability to build trustful
social relationships is an important factor for stimulating cognitive and affective e­ mpathy22.
Aligning with the recent move towards using Research Domain Criteria (RDoC) that aim to provide a bio-
logical framework for understanding mental disorder (based on direct observations rather than just on reports
of symptoms; see e.g., Fonagy and L ­ uyten23), neuroeconomic games such as the “Trust Game” (TG) seem to be
a good paradigm to measure interpersonal t­ rust24. One player (the investor) is endowed with a certain amount
of money (or points that represent a specific monetary value). The investor can decide whether to keep all the
money or to “invest” and to transfer some amount to his partner in the TG (the trustee). The amount invested
is tripled in value and sent to the trustee, who has to decide what proportion to return to the investor. Among
children and adolescents from community samples, two previous studies found a negative association between
externalizing behaviours (conduct and oppositional defiant problems) and investments in a ­TG25,26. However,
a more recent study in a sample of psychiatric inpatients found youth with externalizing behaviour problems
invested less in the first rounds of a repeated T ­ G27. Among male adults, one study reported lower investments in
adult offenders under community supervision compared to non-delinquent ­controls28. Furthermore, outcomes
of a TG depend on the opponent’s characteristics: lower investments were found in TGs against human oppo-
nents only and were not reported in TGs against a computer ­opponent29. The heterogeneity of the samples and
a broader definition of externalizing behaviours may be responsible for the inconsistency of previous findings.
CD related callous-unemotional-traits seem to result from interpersonal deficits in early c­ hildhood23 but their
relation to interpersonal trust is unclear. White et al.30 found fear-specific emotion recognition deficits in children
with high levels of callous-unemotional-traits. Such neurophysiological deficits may lead to poorer and less suc-
cessful social interactions with others and possibly results in lower levels of social trust. In addition, although
several risk assessment instruments exist for predicting criminal recidivism in ­youth31, and socio-emotional
deficits were considered as risk factors in some of them (e.g., Psychopathy Checklist Youth Version)32, no previ-
ous study so far has tested the association of interpersonal trust with any criminal and/or violent re-offenses.
Building on previous research on neuro-economics and psychiatric d ­ isorders33–35, the present study aimed
at analysing a subsample of male juveniles who committed serious violent offenses and met criteria of conduct
disorder (JO/CD) and a subsample of healthy controls (HC) using a series of one-shot trust games (TGs) against
different unknown human opponents or against a computer. Four principal research questions were examined
in this exploratory study: (1) do JO/CD show lower investments than HC? (2) Do investments of human vs.
computer condition differ in TGs? (3) Is overall investment in TGs correlated with callous-unemotional traits
in JO/CD? 4) Is overall investment in TGs associated with any future criminal or violent re-offenses in JO/CD?

Results
Clinical characteristics of the samples.  Information on study participants (e.g., age, foreign national-
ity, living situation), current psychiatric disorders and ICU scores of the total sample and the JO/CD and the
HC subsamples is shown in Table  1. Out of 24 JO/CD, 5 were on psychiatric medication. Two participants
with attention-deficit-hyperactivity disorder were medicated at the time of the assessment. According to clinical
impression and the self-declaration of the participants, none of them was under alcohol or drugs at the time of
the experiments”.

Findings of one‑shot TGs.  Mean investment (of possible 0–10 points) in the first trial against a human was
4.21 points (SD 2.41 points) for the complete sample, 4.82 points (SD 2.43 points) for HC, and 3.48 points (SD
2.26 points) for JO/CD. Mean investment in the first trial against a computer was 4.90 points (SD 3.09 points)
for the complete sample, 5.79 points (SD 3.15 points) for HC, and 4.00 points (SD 2.81 points) for JO/CD. Mean
overall investment of all eight rounds in TGs was 42.89 points (SD 20.06 points) for the complete sample, 48.37
points (SD 19.93 points) for HC, and 37.42 points (SD 19.04 points) for JO/CD. Figure 1 shows mean invest-
ments in trials 1–8 for JO/CD and HC, separately.
The findings of the four tested mixed effect models (with and without interaction terms and with and without
considering a random slope of condition in participants) with subjects as random factor and group and condi-
tion as fixed factors are presented in Table 2. Based on likelihood ratio tests, the addition of an interaction term
(model 2), random slopes for condition (model 3) or both (model 4) did not improve the model fit (compared to
model 1). According to the AIC model 3 and according to BIC model 1 was found to best fitting our data. Based
on these findings and by considering the interpretability, we decided to use model 1 as the most simple solution.

Scientific Reports | (2022) 12:7594 | https://doi.org/10.1038/s41598-022-11777-6 2

Vol:.(1234567890)
www.nature.com/scientificreports/

Variables Total sample JO/CD (n = 24) HC (n = 24) Test ­statisicsa p value

Characteristics of study participants
Age (years) (mean, SD) 17.81 (1.50) 17.71 (1.60) 17.92 (1.41) -0.48 .635
Foreign nationality (n, %) 10 (20.8%) 5 (20.8%) 5 (20.8%) 0.00 1.00
Living at home with both parents 23 (47.9%) 6 (25.0%) 18 (75.0%) 10.10 .001
Out of home placement 7 (14.6%) 5 (20.8%) 2 (8.3%) 1.51 .220
School problems (previous class repetition) 15 (31.3%) 13 (54.2%) 2 (8.3%) – .001
School problems (any truancy in the last year) 13 (27.1%) 9 (37.5%) 4 (16.7%) – 1.93
Current psychiatric disorders
CD 24 (50.0%) 24 (100.0%) 0 (0.0%) 1.00 .000
Substance related disorder 10 (20.8%) 10 (41.7%) 0 (0.0%) 16.63 .001
Affective disorder 3 (6.3%) 3 (12.5%) 0 (0.0%) 3.20 .234
Anxiety disorder 3 (6.3%) 3 (12.5%) 0 (0.0%) 3.20 .234
ADHD 2 (4.2%) 2 (8.3%) 0 (0.0%) 2.09 .489
Any psychiatric diagnosis other than CD 15 (31.3%) 15 (62.5%) 0 (0.0%) 21.82 .000
Callous-unemotional traits
ICU total score (mean, SD) 24.08 (8.41) 27.58 (8.91) 20.58 (6.29) 3.14 .003
ICU callous (mean, SD) 8.17 (4.36) 9.92 (4.92) 6.42 (2.89) 3.01 .004
ICU uncaring (mean, SD) 8.54 (3.77) 9.88 (3.40) 7.21 (3.71) 2.60 .013
ICU unemotional (mean, SD) 7.38 (2.34) 7.79 (2.23) 6.96 (2.42) 1.24 .221

Table 1.  Characteristics of study participants (at baseline), current psychiatric disorders, and callous
unemotional traits of the total sample (N = 48), the JO/CD subsample (n = 24), and the HC subsample
(n = 24). Note: JO/CD = male juveniles who committed serious violent offenses and met criteria of conduct
disorder, HC = Healthy controls, SD = standard deviation, CD = Conduct disorder, ADHD = Attention-Deficit-
Hyperactivity-Disorder, ICU = Inventory of Callous-Unemotional Traits, at-test or χ2-test/Fishers’s exact test.

Figure 1.  Mean investments and standardized errors in TGs 1–8 of JO/CD and HC.

JO/CD invested 0.44 (95% CI − 0.01 to 0.89) CHF (almost 10% of their seed capital) less per round on aver-
age than HC, although the effect was not statistically significant (p = 0.054). In contrast, no effect was found for
the human vs. the computer condition. Age was not found to have a significant effect on investment. Additional
analyses including trial number as additional factor in mixed models 5–8 (supplemental table S1) suggest that
model 7 (without interaction term and including a random intercept and random slope of condition) most accu-
rately fit the data. In this model, and similar to the main results above, JO/CD showed a non-significant trend
to invest less than HC and no differences were found for human vs. the computer condition of TGs. However,
the significant effect of trial number suggests a small increase in investment over time (0.09 CHF per round,
95% CI 0.07–0.12, p < 0.001).

Scientific Reports | (2022) 12:7594 | https://doi.org/10.1038/s41598-022-11777-6 3

Vol.:(0123456789)
 www.nature.com/scientificreports/

Model 1 (including random Model 2 (including random Model 3 (including random Model 4 (including random
intercept) intercept) intercept and random slope)a intercept and random slope)a
S (95% CI), p value S (95% CI), p value S (95% CI), p value S (95% CI), p value
Fixed effects
Group (JO/CD = 1, HC = 0) − 0.44 (− 0.88–0.01), p = .054 − 0.51 (− 0.97–0.04), p = .034) − 0.39 (− 0.83–0.04), p = .075 − 0.51 (− 1.01–0.01), p = .049
Condition (human = 1, com-
− 0.03 (− 0.17–0.10), p = .599 − 0.10 (− 0.29–0.08), p = .273 − 0.03 (− 0.19–0.12), p = .650 − 0.10 (− 0.31–0.11), p = .341
puter = 0)
Group × condition − 0.14 (− 0.12–0.40), p = .305 0.13 (− 0.16–0.44), p = .373
Covariates
Age − 0.01 (− 0.23–0.21), p = .934 − 0.01 (− 0.23–0.21), p = .934 0.01 (− 0.21–0.22), p = .956 0.01 (− 0.21–0.22), p = .956
Model parameters
2-log likelihood − 873.685 − 873.150 − 871.194 − 870.749
AIC 1759.369 1760.301 1758.298 1759.498
BIC 1783.073 1787.955 1789.903 1795.054

Table 2.  Findings from mixed effect models with group and condition as fixed factors and subjects as
random factors on investments. a Including random slope on condition (human = 1, computer = 0), JO/CD
male juveniles who committed serious violent offenses and met criteria of conduct disorder, HC healthy
controls, AIC akaike information criterion, BIC Bayesian information criterion, S standardized coefficient, CI
confidence interval.

Measures 1 2 3 4 5
1 overall investment in TG –
2 ICU total score − .332 (p = .113) –
3 ICU callous − .230 (p = .279) .938 (p < .001) -
4 ICU uncaring − .428 (p = .037) .842 (p < .001) .706 (p < .001) –
5 ICU unemotional − .123 (p = .566) .692 (p < .001) .540 (p = .006) .398 (p = .054) –

Table 3.  Spearman correlations of the overall investment in TG, ICU total score and ICU subscale in JO/CD.
JO/CD = male juveniles who committed serious violent offenses and met criteria of conduct disorder.

Association of overall investment in TGs with self‑reported callous‑unemotional‑traits.  The

Spearman correlation matrix of overall investment in TGs, the ICU total score and subscales ICU-callous, ICU-
uncaring, and ICU-unemotional for JO/CD is shown in Table  3. For the ICU total score, no significant cor-
relation was found between overall investment in TGs. For the subscales a significant negative correlation was
found between overall investment in TGs and the ICU-uncaring but not between overall investment in TGs and
ICU-callous or ICU-unemotional. In further exploratory analyses of the subsample of JO/CD no significant
association was found between overall investment in TGs and any comorbid psychiatric disorders (r = − 0.082,
p = 0.703) and any substance use disorders (r = − 0.209, p = 0.326).

Criminal recidivism.  Of the 24 in the JO/CD subsample, 15 (62.5%) committed one or more criminal re-
offenses and 7 (29.2%) committed one or more violent re-offenses. The proportional hazard assumption was not
found to be violated for overall investment in TGs and age (p > 0.5). Findings of univariate und multiple Cox-
regression with overall investment of TGs as predictor and time to first violent or general re-offense as outcome
variables are shown in Table 4. Model 2 further includes age as covariate in the analyses. Hazard ratios (HRs)
indicate a significant negative association of the overall investment in TGs with later general re-offenses but no
association with later violent re-offenses. JO/CD with smaller overall investments in TGs committed general re-
offenses more frequently in a shorter time period after the TGs. This association was still confirmed when age
was included as covariate in the analyses.

Discussion
The main aim of this study was to examine if JO/CD show lower investments than HC in a series of one-shot
TGs. Mixed model analyses revealed no statistically significant effect between groups (using two tailed testing).
Despite the statistically non-significant result, our data suggest a trend that JO/CD from the forensic outpatient
treatment sample invested less points than HC during TGs, but our sample was too small to detect it (by inter-
preting the point estimate of − 0.44 with 95% confidence intervals of − 0.88 to 0.01, p = 0.054). This result was
also confirmed by additional analyses including trial number as an additional factor in the analyses. Possible
deficits in social trust should be addressed by further research including larger samples and more juveniles with
more severe offenses (e.g., samples from residential care or juvenile detention facilities). Although we included

Scientific Reports | (2022) 12:7594 | https://doi.org/10.1038/s41598-022-11777-6 4

Vol:.(1234567890)
www.nature.com/scientificreports/

Any criminal recidivism Any violent recidivism

Predictors HR (95% CI) p value) HR (95% CI) p value
Model 1
Overall investment in TGs (z-transformed) 0.49 (0.24–0.99) .048 1.09 (0.51–2.36) .820
Model 2
Overall investment in TGs (z-transformed) 0.41 (0.21–0.83) .014 0.92 (0.42–2.03) .838
Age 0.76 (0.54–1.05) .096 0.76 (0.48–1.22) .255

Table 4.  Findings from univariate Cox-regressions with overall investments in TGs as predictors of violent
recidivism and general recidivism in JO/CD subsample (n = 24). Note: male juveniles who committed serious
violent offenses and met criteria of conduct disorder, HR = Hazard Ratio.

age as a covariate in the analyses, which was unrelated to investment, CD may behave differently in younger
compared to older aged adolescents.
Although, the group effect closely missed significance our findings mirror previous findings of adults who
were under supervision because of criminal offenses and were found to invest less in a one shot TG against an
unknown opponent than community c­ ontrols28. However, in contrast to the study of Clark, et al.28, we exam-
ined a series of one-shot TGs involving different human and computer opponents and found consistently lower
investments in JO/CD for each trial. Previous findings only partly confirm trust deficits at the beginning of an
interaction with specific human opponents in ­youths25–27. Their findings were based on a broader definition of
externalizing behaviours and the inclusion of younger children of both genders.
Juvenile antisocial behaviours with lower social investments in interactions seem to reflect a general social
mistrust and a “hostility bias”20 that may go along with previous and current social experiences of adolescents
and young adults with antisocial behaviours. Offenders tend to believe in an unjust ­world36. A number of studies
have found an accumulation of adverse childhood experiences (ACE; including e.g., physical childhood maltreat-
ment, neglect, parental separation) to be strongly associated with the development of CD and current/future
violent behaviours in adolescence and young a­ dulthood37,38. In line with this research, a recent study among
incarcerated males found interpersonal trust deficits as a consequence of an inability to form moral impression
on harmfulness of others after the exposure to v­ iolence39. However, although the general strategy used by the
JO/CD is not the optimal strategy under trustful conditions, it may be the optimal strategy given the rules of
a criminal and egoistic world. In summary, our findings suggest that social trust is an interesting concept that
should be addressed further to understand the development of aggressive and violent behaviours.
Our second research question relates to additional effects of human vs. computer condition on investments
in TGs. Investments in TGs did not differ between the human and computer conditions. Previous studies sug-
­ umans27. In a lottery condition where outcomes
gest that trust deficits were specific to social interactions with h
were randomly selected by a computer algorithm from a range of possible outcomes no differences were found
between youth with externalizing behaviour problems and c­ ontrols27. Such a lottery setting is particularly related
to risk taking decisions that were found to be higher in children and adolescents with A ­ DHD40, that is also fre-
quently comorbid with C ­ D2. However, the computer condition in the current study is more comparable with a
real social interaction because the algorithm randomly selected a previous response from a database of human
opponents (in order to warrant similar returns and to produce realistic responses). Therefore, the fact alone that
the participants knew that a computer algorithm makes the decision seems not to be relevant for decision making.
Probably, the participants expectations of higher returns by computers becomes of importance in lottery games.
In additional analyses including trial number as predictor in mixed models (see supplemental table S1) we
found a significant effect of time course of trials. This finding was rather unexpected as in the present study par-
ticipants played TGs against different unknown human opponents and it was not possible to learn from previous
interactions. Our findings mirror studies of repeated TGs that were played against the same human opponent.
In such settings trust emerges as the outcome of a reciprocal social ­relationship41. We have no final explanation
for this finding. It can be assumed that trust may also arise from unspecific experiences of social success or that
some youth behave like they were playing a repeated TG even they know that this was not the case. The increase
of investments in JO/CD further suggest that also antisocial youths are able to increase social trust but may need
more time in a trustful environment than healthy subjects. This may be important for interventions with JO/CD.
Further analyses in the subsample of JO/CD focused on the association of trust with self-reported callous-
unemotional traits. No association of the ICU total score as well as of the ICU-callousness and ICU-unemotional
with overall investments was found in JO/CD. Hence, our findings do not support the idea that general callous-
unemotional traits or more specific emotional dysfunctions such as a lack of empathy and remorse or emotional
perception were relevant for JO/CD’s performance in one-shot TGs. However, a significant negative correlation
of − 0.428 between the ICU-uncaring and overall investment in TGs in JO/CD was detected. JO/CD with higher
investments were found to be less uncaring about others. This finding supports the relevance of trust related
abnormalities in the moral development in some adolescents with C ­ D5 that may linked to early deficits in social
bonding to ­caregivers42. Further exploratory analyses revealed no association of total investment in TGs with
any comorbid psychiatric disorders/comorbid substance use disorders. Substance use disorders are well-known
comorbidities of C ­ D43 and highly prevalent in juvenile detention s­ amples44.
Finally, we tested trust as predictor of criminal recidivism. Findings from univariate Cox-regressions revealed
that overall investment in TGs is a significant predictor of general but not violent recidivism. The more JO/CD
invested in TGs, the lower was the risk for committing any further offense within the next 2.9–7 years. Although

Scientific Reports | (2022) 12:7594 | https://doi.org/10.1038/s41598-022-11777-6 5

Vol.:(0123456789)
 www.nature.com/scientificreports/

this finding is based on a small sample of 24 JO/CD, it suggests a potential for TG as a prognostic tool in forensic
psychiatry/psychology. Interpersonal trust seems not only related to current antisocial behaviours but also to
have long lasting effects on criminal behaviours. This result further supports the concept of interpersonal trust
to understand social behaviour in adolescents and young adults charged or convicted for violent offenses. To
the best of our knowledge, this study is the first study to examine the effects of investments in TGs on criminal
recidivism. Future studies may address this issue within larger samples of JO/CD subjects, controlling for well-
established prognostic factors such as criminal history, psychopathology and personality traits.
The limitations of the study include the following: the current findings are based on a rather small sample of
one forensic institution and aged matched controls. The small sample size and the focus on males limit the gener-
alizability of the findings. Further replications in severely offending juveniles (e.g., from residential care, juvenile
detention centres) and larger samples seem necessary. Although HC were interviewed to exclude psychiatric
disorders and (serious) criminal activities no further validation of their declarations was possible. No informa-
tion on participants socioeconomic background and on adverse childhood experiences was available. Both were
previously reported as a risk factor for criminal offenses and r­ ecidivism37,45,46. Furthermore, no information on
the severity, the beginning and the duration of CD and the presence of antisocial personality disorder (for JO/CD
older than 18 years) was available. Although IQ < 80 was considered as exclusion criteria in both subsamples, we
cannot rule out that IQ and other factors may affected the findings of the TGs. The definition of trust as invest-
ment in TGs may differ from trust-based decisions in daily life of adolescents. Future studies may benefit from
the inclusion of a questionnaire addressing relational trust. Furthermore, the limited statistical power impeded
the performance of further analyses on covariates. Considering only official reoffending data, we cannot rule
out that some re-offenses may not have been reported to the police or were not charged by justice authorities.
The current finding on social trust deficits in CD suggests that trust may be an interesting concept for forensic
psychiatry and may impact further research. In addition, it suggests that measuring social behaviours in real
time during game play may offer new possibilities for forensic assessments. Given the tendencies of juveniles to
deny or minimize problem behaviours in questionnaires and interviews, behavioural economic approaches may
serve as additional possibilities to study underlying mechanisms of CD symptoms, to estimate the risk of further
offending and to tailor interventions. The use of self-report questionnaires to assess externalizing behaviours is
limited, because they refer to the participants general knowledge on socially acceptable behaviours rather than
revealing the participants actual feelings and behaviours during social interactions. Our results suggest that
interpersonal trust may help to understand antisocial behaviours in individuals with JO/CD and may be used
to predict general criminal recidivism. In addition, our findings of increasing trust in JO/CD individuals during
TGs 1–8 suggests trust may represent a target for therapeutic intervention for CD.

Methods
Participants.  Juveniles who committed serious violent offenses and met criteria of conduct disorder (JO/
CD).  Data were collected between October 2013 and December 2017 from male adolescents referred by ju-
venile justice authorities for a forensic treatment at the Child and Adolescent Forensic Outpatient Unit of the
University Hospital of Psychiatry Zurich. Inclusion criteria were the presence of a conduct disorder (CD; accord-
ing to MINI-KID criteria see below), male gender, age of 14–21 years, and the willingness of the participation
in the current study provided by an informed consent of the youth and a parent (for adolescents below age of
18 years). Exclusion criteria were any current psychotic disorder, any current substance dependence, any current
or past severe eating disorder, any current depressive or manic episode and a severe attention-deficit disorder
(according to MINI-KID and clinical impression of the therapists) as well as an IQ < 80 (according to clinical im-
pression). The final JO/CD sample consisted of 24 male participants (mean age 17.71 years, SD 1.60 years, range
14–20 years). All of them were adjudicated or convicted of having committed serious violent offenses according
to the Swiss penal law (e.g., violent assault, robbery, sexual offenses). Eligible probands were approached by their
therapists. Subsequently, they have been invited personally to participate in the study by the first author (MA)
who worked at the same unit and joined at the beginning of a therapy session. After agreeing to participate, fur-
ther appointments were arranged with the independent experimenter (MH). Confidentiality was assured for all
information that was collected in the study.

Healthy controls (HC).  Data were collected between January 2014 and October 2017 on male adolescents
recruited via flyers in schools. After responding they were directly invited by the experimenter (MH). Inclusion
criteria were comparable to JO/CD. Exclusion criteria were the presence of any current psychiatric disorder
(according to MINI-KID including CD), an IQ < 80 (according to clinical impression) and the presence of crimi-
nal activities or convictions (according to interviews with HC). The final HC sample was matched on age and
consisted of 24 male participants (mean age 17.92 years, SD 1.41 years, range 16–21 years).

Measures.  Conduct disorder (CD)/psychiatric disorders.  The presence of a psychiatric disorder (including
CD and comorbidity) was assessed via the Mini-International Neuropsychiatric Interview for Children and
Adolescents (MINI-KID), a short structured clinical diagnostic interview designed to assess the presence of
psychiatric disorders in children and a­ dolescents47. Diagnoses were based on an algorithm that is appropriate
for symptom count, age, duration, and impairment according to DSM-IV criteria. We assessed the presence of
conduct disorders (CD; inclusion criteria for JO/CD) as well as substance related disorders, affective disorders
(e.g., major depression, dysthymia), anxiety disorders (e.g., social phobia, separation anxiety, panic disorders),
and attention-deficit-hyperactivity-disorder. The MINI-KID has been shown to have good reliability and valid-
ity when compared with the Schedule for Affective Disorders and Schizophrenia for School-Aged Children—
Present and Lifetime ­Version47.

Scientific Reports | (2022) 12:7594 | https://doi.org/10.1038/s41598-022-11777-6 6

Vol:.(1234567890)
www.nature.com/scientificreports/

Callous‑unemotional traits.  The self-report version of the Inventory of Callous-Unemotional traits [­ ICU48] was
used as a dimensional measure to study callous-unemotional traits in JO/CD and HC. The scale consists of 24
items using a 4-point Likert scale from 0 (not at all true) to 3 (absolutely true). Factor analyses confirmed a gen-
eral ICU total score and three distinct factors: callousness (e.g., “The feelings of others are unimportant to me”),
uncaring (e.g., “I try not to hurt others’ feelings”, reversed score), and unemotional (e.g., “It is easy for others to
tell how I am feeling”, reversed score). The ICU was found to be a reliable and valid measure in adolescents (using
the German version of the ICU) and juvenile ­offenders49,50. For the present study the ICU total score and the
subscales were used (Cronbach alpha: ICU total score = 0.822, ICU-callousness = 0.730, ICU-uncaring = 0.707,
ICU-unemotional = 0.524). No official cut-off scores are available for the ICU. A recent ­study51 analysed possible
cut-off scores in a community and a juvenile detention sample from the US and suggested an empirical based
cut-off score of 24.

Criminal recidivism.  Information on criminal recidivism after TGs of the 24 JO/CD (mean time 5.5 years, SD
1.14 years, time range 2.91–6.98 years) were drawn on November 18, 2020 from official records by the Swiss Fed-
eral Office of Justice. In line with previous studies [e.g.,52] on juvenile offending we first analyzed (time to) any
criminal re-offense. In a second step, we analyzed (time to) violent re-offenses as an indicator for more serious
offenses. Violent crimes were considered as both general reoffences and violent re-offences. Any criminal recidi-
vism was coded according to any conviction after TGs or any current adjudication on crimes as defined by the
Swiss Penal Code. Violent recidivism was coded based on any new conviction or current adjudication on crimes
as defined by the section “crimes against body and life” in the Swiss Penal Code (e.g., murder, manslaughter,
bodily assault) as well as robbery, coercion, and any contact sexual offenses (rape, sexual coercion). Besides the
presence of any new (violent) criminal conviction or adjudication we also coded time to the first re-offense and
first violent re-offense after the assessment based on the offending dates reported in files.

Procedure.  One‑shot trust games.  Each participant played a series of eight one-shot trust games (TGs). The
participants acted as investors against a recipient/trustee, either a human player (trial 1, 2, 6, and 8) or a com-
puter (trial 3, 4, 5, and 7). In each TG, players have ten points as seed capital (each point represents a value of 0.5
Swiss francs). The investor is now asked to transfer an amount × ranging from 0 to 10 points from their capital
to the recipient. This amount is tripled, common for trust games (***Camerer, 2003). The recipient/trustee can
now decide, how many points he wants to return to the initial investor (ranging from 0 points to the possible
maximum of his points available). The participants were immediately informed of the amount that was returned
from the trustees.
In the human condition, the role of the opposing player was played by the investigator (MH), using the
responses of trustees who had played previously. Participants randomly choose one opponent for each trial out
of a pool of n = 76 youths who have played the game previously recording all their possible reactions based on the
prior investment from 0 to 10. In the computer condition participants were informed that the amount returned
from the trustee was based on the decision of a computer algorithm randomly selecting from all previous trustee
responses. Before the beginning of the testing, TGs were carefully explained to all participants and examples
of a TG against a human player and of a TG against a computer were performed to see if they understood the
procedures. In the human player condition, it was clear for participants that the decisions were made by an
adolescent who had played the game before. After completing the TGs, one trial of the human condition (round
1, 2, 6, or 8) was randomly selected for pay off.

Statistical analyses.  Mixed models were used to analyse main and interaction effects of group (JO/CD vs. HC)
and condition (human vs. computer) on investments in one shot TGs (trial 1–8). Analyses were performed in R
Statistical Software (version 3.6.3) using packages “nlme”53 using maximum likelihood estimations. We fit four
models to the investment data. Model 1 contained group and condition as independent factors. Model 2 added
the interaction between group and condition. To control for interindividual variations, subjects were defined as a
random factor (by including a random intercept in the analyses) in all models. In models 3 (without interaction)
and 4 (with interaction), we additionally considered a random slope on condition (human vs. computer). Age
was considered as covariate in all analyses. In order to choose a single model to evaluate, we used likelihood ratio
testing for nested models, the Akaike Information Criterion [­AIC54] and the Bayesian Information Criterion
­[BIC55], where lower AIC or BIC values indicate a better fit to the data. Although we did not use a repeated trust
game, where participants played several rounds against the same opponent, we cannot exclude the presence of
a time effect between trials 1–8. To test this, we repeated the mixed model analyses that we performed before
by including trial number as an additional independent factor in the analysis. For all mixed models, standard-
ized coefficients and 95% confidence intervals were included to compare effects (see discussion on effect sizes
in mixed models)56.
Spearman correlations were used to analyse associations of overall investment in TGs and ICU total score and
subscales in JO/CD. Cox-regressions were used to analyse the association of overall investment in TGs with time
to any criminal recidivism and time to any violent re-offenses. The Cox-proportional hazard model is essentially
a regression model commonly used in medical research for investigating the association between the survival
time of patients and one or more predictor variables. In the present study, the time to the first criminal re-offense
was considered as survival time (dependent variable) and overall investment in TGs as independent variable
(model 1). Based on the well-known association of young age and increased risk for criminal (re-)offenses57,58,
age was as included as a covariate (model 2). The hazard ratio from Cox regression can be interpreted as relative
instantaneous ­risk59. It indicates the change in the risk of an event (e.g., a criminal re-offense) if the parameter
in question (e.g., overall investment in TGs) rises by one unit. A hazard ratio of one means lack of association,

Scientific Reports | (2022) 12:7594 | https://doi.org/10.1038/s41598-022-11777-6 7

Vol.:(0123456789)
 www.nature.com/scientificreports/

a hazard ratio greater than one suggests an increased risk, and a hazard ratio below one suggests a smaller risk.
Statistically significant hazard ratios cannot include unity (one) in their confidence intervals. The proportional
hazard assumption (it essentially means is that the ratio of the hazards for any two individuals is constant over
time) was tested for overall in TGs and age.
Analyses were performed in SPSS (version 24) and R statistic software version 3.6.360. The level of significance
was set to be p < 0.05 (two tailed).

Ethics approval.  Ethical approval was granted by the ethics committee of the Canton Zurich (Ref. no. 2012-
0438). This study was performed in accordance with the ethical standards as laid down in the 1964 Declaration
of Helsinki and its later amendments or comparable ethical standards.

Informed consent.  Informed consent was obtained from all participants and a legal guardian (for adoles-
cents below age of 18 years).

Data availability
Data are available on request by the corresponding author.

Received: 1 November 2021; Accepted: 1 April 2022

References
1. World Health Organization. Preventing Youth Violence: An Overview of the Evidence (World Health Organization, 2015).
2. Aebi, M., Barra, S., Bessler, C., Walitza, S. & Plattner, B. The validity of conduct disorder symptom profiles in high-risk male youth.
Eur. Child Adolesc. Psychiatry 28, 1537–1546. https://​doi.​org/​10.​1007/​s00787-​019-​01339-z (2019).
3. Fairchild, G. et al. Conduct disorder. Nat. Rev. Dis. Primers 5, 1–25. https://​doi.​org/​10.​1038/​s41572-​019-​0095-y (2019).
4. Frick, P. J., Ray, J. V., Thornton, L. C. & Kahn, R. E. Can callous-unemotional traits enhance the understanding, diagnosis, and
treatment of serious conduct problems in children and adolescents? A comprehensive review. Psychol. Bull. 140, 1. https://​doi.​
org/​10.​1037/​a0033​076 (2014).
5. Brugman, D., Out, C. & Gibbs, J. C. In Women and Children as Victims and Offenders: Background, Prevention, Reintegration:
Suggestions for Succeeding Generations (Volume 1) (eds Kury, H. et al.) 265–289 (Springer, 2016).
6. Krueger, F. & Meyer-Lindenberg, A. Toward a model of interpersonal trust drawn from neuroscience, psychology, and economics.
Trends Neurosci. 42, 92–101. https://​doi.​org/​10.​1016/j.​tins.​2018.​10.​004 (2019).
7. Rousseau, D. M., Sitkin, S. B., Burt, R. S. & Camerer, C. Not so different after all: A cross-discipline view of trust. Acad. Manage.
Rev. 23, 393–404. https://​doi.​org/​10.​5465/​amr.​1998.​926617 (1998).
8. Erikson, E. H. Childhood and Society (WW Norton & Company, 1963).
9. Bowlby, J. Attachment and loss: Retrospect and prospect. Am. J. Orthopsychiatry 52, 664. https://​doi.​org/​10.​1111/j.​1939-​0025.​
1982.​tb014​56.x (1982).
10. Sutter, M. & Kocher, M. G. Trust and trustworthiness across different age groups. Games Econ. Behav. 59, 364–382. https://​doi.​
org/​10.​1016/j.​geb.​2006.​07.​006 (2007).
11. Hepp, J., Schmitz, S. E., Urbild, J., Zauner, K. & Niedtfeld, I. Childhood maltreatment is associated with distrust and negatively
biased emotion processing. Borderline Personal. Disord. Emot. Dysregul. 8, 1–14. https://​doi.​org/​10.​1186/​s40479-​020-​00143-5
(2021).
12. Bell, V., Robinson, B., Katona, C., Fett, A.-K. & Shergill, S. When trust is lost: The impact of interpersonal trauma on social interac-
tions. Psychol. Med. 49, 1041–1046. https://​doi.​org/​10.​1017/​S0033​29171​80018​00 (2019).
13. Fett, A.-K.J. et al. To trust or not to trust: The dynamics of social interaction in psychosis. Brain 135, 976–984. https://​doi.​org/​10.​
1093/​brain/​awr359 (2012).
14. Poggi, A., Richetin, J. & Preti, E. Trust and rejection sensitivity in personality disorders. Curr. Psychiatry Rep. 21, 69. https://​doi.​
org/​10.​1007/​s11920-​019-​1059-3 (2019).
15. Stams, G. J. et al. The moral judgment of juvenile delinquents: A meta-analysis. J. Abnorm. Child Psychol. 34, 692–708. https://​doi.​
org/​10.​1007/​s10802-​006-​9056-5 (2006).
16. Van Vugt, E. et al. Moral development and recidivism: A meta-analysis. Int. J. Offender Ther. Comp. Criminol. 55, 1234–1250.
https://​doi.​org/​10.​1177/​03066​24X11​396441 (2011).
17. Harms, M. B., Leitzke, B. T. & Pollak, S. D. Handbook of Emotional Development 767–786 (Springer, 2019).
18. Sebastian, C. L. et al. Modulation of amygdala response by cognitive conflict in adolescents with conduct problems and varying
levels of CU traits. Res. on Child and Adolesc. Psychopath. 49, 1043–1054. https://​doi.​org/​10.​1007/​s10802-​021-​00787-z (2021).
19. White, S. F. et al. Exposure to violence and low family income are associated with heightened amygdala responsiveness to threat
among adolescents. Dev. Cogn. Neurosci. 40, 100709. https://​doi.​org/​10.​1016/j.​dcn.​2019.​100709 (2019).
20. Crick, N. R. & Dodge, K. A. Social information-processing mechanisms in reactive and proactive aggression. Child Dev. 67,
993–1002. https://​doi.​org/​10.​2307/​11318​75 (1996).
21. De Castro, B. O., Veerman, J. W., Koops, W., Bosch, J. D. & Monshouwer, H. J. Hostile attribution of intent and aggressive behavior:
A meta-analysis. Child Dev. 73, 916–934. https://​doi.​org/​10.​1111/​1467-​8624.​00447 (2002).
22. Rotenberg, K. J. & Mann, L. The development of the norm of the reciprocity of self-disclosure and its function in children’s attrac-
tion to peers. Child Dev. https://​doi.​org/​10.​2307/​11304​14 (1986).
23. Fonagy, P. & Luyten, P. Conduct problems in youth and the RDoC approach: A developmental, evolutionary-based view. Clin.
Psychol. Rev. 64, 57–76. https://​doi.​org/​10.​1016/j.​cpr.​2017.​08.​010 (2018).
24. Berg, J., Dickhaut, J. & McCabe, K. Trust, reciprocity, and social history. Games Econ. Behav. 10, 122–142. https://​doi.​org/​10.​1006/​
game.​1995.​1027 (1995).
25. Sharp, C., Burton, P. C. & Ha, C. “Better the devil you know”: A preliminary study of the differential modulating effects of reputa-
tion on reward processing for boys with and without externalizing behavior problems. Eur. Child Adolesc. Psychiatry 20, 581–592.
https://​doi.​org/​10.​1007/​s00787-​011-​0225-x (2011).
26. Sharp, C., Carolyn, H. & Fonagy, P. Get them before they get you: Trust, trustworthiness, and social cognition in boys with and
without externalizing behavior problems. Dev. Psychopathol. 23, 647–658. https://​doi.​org/​10.​1017/​S0954​57941​00000​03 (2011).
27. Mellick, W., Sharp, C. & Sumlin, E. Trust and general risk-taking in externalizing adolescent inpatients versus non-externalizing
psychiatric controls. Scand. J. Child Adolesc. Psychiatr. Psychol. 7, 92–96. https://​doi.​org/​10.​21307/​sjcapp-​2019-​013 (2019).
28. Clark, B. C. et al. Individuals in the criminal justice system show differences in cooperative behaviour: Implications from coopera-
tive games. Crim. Behav. Ment. Health 25, 169–180. https://​doi.​org/​10.​1002/​cbm.​1920 (2015).

Scientific Reports | (2022) 12:7594 | https://doi.org/10.1038/s41598-022-11777-6 8

Vol:.(1234567890)
www.nature.com/scientificreports/

29. Tzieropoulos, H. The Trust Game in neuroscience: A short review. Soc. Neurosci. 8, 407–416. https://​doi.​org/​10.​1080/​17470​919.​
2013.​832375 (2013).
30. White, S. F. et al. Can the fear recognition deficits associated with callous-unemotional traits be identified in early childhood?. J.
Clin. Exp. Neuropsychol. 38, 672–684. https://​doi.​org/​10.​1080/​13803​395.​2016.​11491​54 (2016).
31. Schwalbe, C. S. Risk assessment for juvenile justice: A meta-analysis. Law Hum. Behav. 31, 449. https://​doi.​org/​10.​1007/​s10979-​
006-​9071-7 (2007).
32. Jeon, H., Boccaccini, M. T., Jo, E., Jang, H. & Murrie, D. C. Rater experience and the predictive validity of Psychopathy Checklist:
Youth Version scores. Psychiatr. Psychol. Law https://​doi.​org/​10.​1080/​13218​719.​2020.​17513​30 (2020).
33. Hasler, G. Can the neuroeconomics revolution revolutionize psychiatry?. Neurosci. Biobehav. Rev. 36, 64–78. https://​doi.​org/​10.​
1016/j.​neubi​orev.​2011.​04.​011 (2012).
34. King-Casas, B. & Chiu, P. H. Understanding interpersonal function in psychiatric illness through multiplayer economic games.
Biol. Psychiatry 72, 119–125. https://​doi.​org/​10.​1016/j.​biops​ych.​2012.​03.​033 (2012).
35. Preuss, N. et al. Inconsistency and social decision making in patients with Borderline Personality Disorder. Psychiatry Res. 243,
115–122. https://​doi.​org/​10.​1016/j.​psych​res.​2016.​06.​017 (2016).
36. Walters, G. D. Correlations between the Psychological Inventory of Criminal Thinking Styles and World-View Rating Scale in
male federal prisoners. Crim. Behav. Ment. Health 17, 184–188. https://​doi.​org/​10.​1002/​cbm.​649 (2007).
37. Baglivio, M. T., Wolff, K. T., Piquero, A. R. & Epps, N. The relationship between adverse childhood experiences (ACE) and juvenile
offending trajectories in a juvenile offender sample. J. Crim. Justice 43, 229–241. https://​doi.​org/​10.​1016/j.​jcrim​jus.​2015.​04.​012
(2015).
38. Craig, J. M., Piquero, A. R., Farrington, D. P. & Ttofi, M. M. A little early risk goes a long bad way: Adverse childhood experiences
and life-course offending in the Cambridge study. J. Crim. Justice 53, 34–45. https://d ​ oi.​org/1​ 0.1​ 016/j.j​ crimj​ us.2​ 017.0​ 9.0​ 05 (2017).
39. Siegel, J. Z., Estrada, S., Crockett, M. J. & Baskin-Sommers, A. Exposure to violence affects the development of moral impressions
and trust behavior in incarcerated males. Nat. Commun. 10, 1–9. https://​doi.​org/​10.​1038/​s41467-​019-​09962-9 (2019).
40. Pollak, Y., Dekkers, T. J., Shoham, R. & Huizenga, H. M. Risk-taking behavior in attention deficit/hyperactivity disorder (ADHD):
A review of potential underlying mechanisms and of interventions. Curr. Psychiatry Rep. 21, 1–11. https://d ​ oi.o​ rg/1​ 0.1​ 007/s​ 11920-​
019-​1019-y (2019).
41. Cochard, F., Van, P. N. & Willinger, M. Trusting behavior in a repeated investment game. J. Econ. Behav. Organ. 55, 31–44. https://​
doi.​org/​10.​1016/j.​jebo.​2003.​07.​004 (2004).
42. Dadds, M. R., English, T., Wimalaweera, S., Schollar-Root, O. & Hawes, D. J. Can reciprocated parent–child eye gaze and emotional
engagement enhance treatment for children with conduct problems and callous-unemotional traits: A proof-of-concept trial. J.
Child Psychol. Psychiat. 60, 676–685. https://​doi.​org/​10.​1111/​jcpp.​13023 (2019).
43. Frick, P. J. Current research on condurct disorders in children and adolescents. S. Afr. J. Psychol. 46, 160–174. https://​doi.​org/​10.​
1177/​00812​46316​628455 (2016).
44. Colins, O. et al. Psychiatric disorders in detained male adolescents: A systematic literature review. Can. J. Psychiat. 55, 255–263.
https://​doi.​org/​10.​1177/​07067​43710​05500​409 (2010).
45. Aebi, M., Giger, J., Plattner, B., Metzke, C. W. & Steinhausen, H.-C. Problem coping skills, psychosocial adversities and mental
health problems in children and adolescents as predictors of criminal outcomes in young adulthood. Eur. Child Adolesc. Psychiatry
23, 283–293. https://​doi.​org/​10.​1007/​s00787-​013-​0458-y (2014).
46. Cottle, C. C., Lee, R. J. & Heilbrun, K. The prediction of criminal recidivism in juveniles: A meta-analysis. Crim. Justice Behav. 28,
367–394. https://​doi.​org/​10.​1177/​00938​54801​02800​3005 (2001).
47. Sheehan, D. V. et al. Reliability and validity of the mini international neuropsychiatric interview for children and adolescents
(MINI-KID). J. Clin. Psychiat. 71, 313–326. https://​doi.​org/​10.​4088/​JCP.​09m05​305whi (2010).
48. Frick, P. J. The Inventory of Callous-Unemotional Traits (University of Luisiana, 2004).
49. Essau, C. A., Sasagawa, S. & Frick, P. J. Callous-unemotional traits in a community sample of adolescents. Assessment 13, 454–469.
https://​doi.​org/​10.​1177/​10731​91106​287354 (2006).
50. Kimonis, E. R. et al. Assessing callous–unemotional traits in adolescent offenders: Validation of the Inventory of Callous-Unemo-
tional Traits. Int. J. Law Psychiatry 31, 241–252. https://​doi.​org/​10.​1016/j.​ijlp.​2008.​04.​002 (2008).
51. Kemp, E. C. et al. Developing cutoff scores for the inventory of callous-unemotional traits (ICU) in justice-involved and community
samples. J. Clin. Child Adolesc. Psychol. https://​doi.​org/​10.​1080/​15374​416.​2021.​19553​71 (2021).
52. Barra, S. et al. Adverse childhood experiences, personality, and crime: Distinct associations among a high-risk sample of institu-
tionalized youth. Int. J. Environ. Res. Public Health 19, 1227. https://​doi.​org/​10.​3390/​ijerp​h1903​1227 (2022).
53. Pinheiro, J. et al. Package ‘nlme’. Linear and Nonlinear Mixed Effects Models, Version 3 (2017).
54. Akaike, H. A new look at the statistical model identification. IEEE Trans. Autom. Control 19, 716–723. https://​doi.​org/​10.​1109/​
TAC.​1974.​11007​05 (1974).
55. Schwarz, G. Estimating the dimension of a model. Ann. Stat. 6, 461–464 (1978).
56. Lorah, J. Effect size measures for multilevel models: Definition, interpretation, and TIMSS example. Large-Scale Assess. Educ. 6,
1–11. https://​doi.​org/​10.​1186/​s40536-​018-​0061-2 (2018).
57. Farrington, D. P. Age and crime. J. Crime Justice 7, 189–250. https://​doi.​org/​10.​1086/​449114 (1986).
58. Moffitt, T. E. Adolescence-limited and life-course-persistent antisocial behavior: A developmental taxonomy. Psychol. Rev. 100,
674. https://​doi.​org/​10.​1037/​0033-​295X.​100.4.​674 (1993).
59. De Neve, J. & Gerds, T. A. On the interpretation of the hazard ratio in Cox regression. Biom. J. 62, 742–750. https://​doi.​org/​10.​
1002/​bimj.​20180​0255 (2020).
60. R Development Core Team. R: A Language and Environment for Statistical Computing (R Foundation for Statistical Computing,
2020).

Acknowledgements
The authors are grateful to Yannick Rothacher from the Institute of Psychology, University of Zurich for his
statistical advice and to Alan Haynes for the English editing of the manuscript.

Author contributions
M.A., M.H., C.B. and G.H. contributed to the study conception and design. Material preparation, data collec-
tion and analysis were performed by M.A. and M.H. The first draft of the manuscript was written by M.A. All
authors commented on previous versions of the manuscript. All authors read and approved the final manuscript.

Funding
This study did not receive external funding.

Scientific Reports | (2022) 12:7594 | https://doi.org/10.1038/s41598-022-11777-6 9

Vol.:(0123456789)
 www.nature.com/scientificreports/

Competing interests 
The authors declare no competing interests.

Additional information
Supplementary Information The online version contains supplementary material available at https://​doi.​org/​
10.​1038/​s41598-​022-​11777-6.
Correspondence and requests for materials should be addressed to M.A.
Reprints and permissions information is available at www.nature.com/reprints.
Publisher’s note  Springer Nature remains neutral with regard to jurisdictional claims in published maps and
institutional affiliations.
Open Access  This article is licensed under a Creative Commons Attribution 4.0 International
License, which permits use, sharing, adaptation, distribution and reproduction in any medium or
format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the
Creative Commons licence, and indicate if changes were made. The images or other third party material in this
article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the
material. If material is not included in the article’s Creative Commons licence and your intended use is not
permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from
the copyright holder. To view a copy of this licence, visit http://​creat​iveco​mmons.​org/​licen​ses/​by/4.​0/.

© The Author(s) 2022

Scientific Reports | (2022) 12:7594 | https://doi.org/10.1038/s41598-022-11777-6 10

Vol:.(1234567890)